The Milnor fiber and the zeta function of the singularities of type f = P

Transcription

The Milnor fiber and the zeta function of the singularities of type f = P
C OMPOSITIO M ATHEMATICA
A NDRÁS N ÉMETHI
The Milnor fiber and the zeta function of the
singularities of type f = P(h, g)
Compositio Mathematica, tome 79, no 1 (1991), p. 63-97.
<http://www.numdam.org/item?id=CM_1991__79_1_63_0>
© Foundation Compositio Mathematica, 1991, tous droits réservés.
L’accès aux archives de la revue « Compositio Mathematica » (http:
//http://www.compositio.nl/) implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/legal.php). Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme
Numérisation de documents anciens mathématiques
http://www.numdam.org/
63
1991.
Compositio Mathematica 79: 63-97,
e 1991 Kluwer Academic Publishers. Printed
in the Netherlands.
The MiInor fiber and the zeta function of the
singularities
of type
f P(h,g)
=
ANDRÁS NÉMETHI
Institute
of Mathematics of Romanian Academy,
Received 24 October 1989;
accepted
20
July
Bdul Pâcii 220, 79622 Bucharest, Romania
1990
1. Introduction
An important invariant of a germ of analytic function f : (C"’ 1, 0)~(C, 0) is its
Milnor fiber F. Milnor [21] proved that if f has an isolated singularity at the
origin then the homotopy type of the fiber is a bouquet of 1À spheres of dimension
n, where J1 is the Milnor number of f. For isolated line singularities (not of type
A~) Siersma [29] proved that the Milnor fiber is homotopy equivalent to a
bouquet of (# A1 +2 # D~-1)-spheres of dimension n, where # A1 and * D.
are the number of A1 and, respectively D~ points in a genéric approximation of
f. More generally, if the critical set Sing V(f) of f is an 1-dimensional complete
intersection (ICIS) and the transversal singularity of f in the points of Sing
V(f) - {0} is of type A1, then in the general ( =tF D 00 &#x3E; 0) case F is homotopic to
VS nand in the special case (# D~ = 0) F is homotopic to Sn-1 v (yS n) [31]. In the
case of line singularities, but with transversal type S = A1, A2, A3, D4, E6, E7, E8
de Jong [37] proved that in the general case the Milnor fiber is homotopic to
(Sn-1) (Sn). Other important cases with 1-dimensional singular locus were
studied by Iomdin [14], Pellikaan [23] and Siersma [32].
Perhaps the most efficient method in the study of the Milnor fiber of nonisolated singularities is the polar slice technique which materialized in two
beautiful formulae for the Euler-characteristic of the fibers: the Lê attaching
formula [16] and the Iomdin-Lê formula [14,18]. Moreover, Lê [17] used the
polar slice construction to show that the geometric monodromy of f preserves
the polar filtration and can be constructed as a carrousel.
In the present paper our main object of study is an analytic germ with
the following property: f can be written as f P(h, g), where the pair
(h, g) : (Cn + B 0) (C2, 0) defines an ICIS and P is a germ of two variables.
Note that if P has a (proper) isolated singularity at the origin then
dim Sing V(f) = n - 1 and if P is not reduced, then dim Sing V(f) n. It is easy
to see that in the latter case even the connectivity of the fiber may fail. But even if
the fiber is connected we can not expect its simple-connectivity because of a
theorem due to Kato and Matsumoto [ 15] giving, in general, the best bound for
=
~
=
64
the connectivity of the fiber
somewhat surprisingly, by
(in terms of the dimension of the singular set). But,
our results we do have a "depth connectivity":
03C0k(F) = 0 for 2 k n - 2.
This is perhaps the first example of a class of germs with non-separable
variables for which we have important informations on the other side of the
"Kato-Matsumoto bound".
More precisely, we prove the
following
THEOREM A.
(a) The Milnor fiber F has the homotopy type of a space obtained from the total
space of a fiber bundle (with base space the.fiber of P and with fiber the general
fiber of the ICIS) by attaching (y f (c), f)0 cells of dimension n.
(b) If P - 1 (0) D {0} (D denotes the reduced discriminant locus of the ICIS)
then F has the homotopy type of 03C00 (Milnor fiber of P) copies of a bouquet of
spheres NS1) v (VS "). (See 2.0.1 and 2.0.2 for the definition of 03B3f(c)).
=
The number of (1-dimensional) circles and n-dimensional spheres in a
connected component is determined in terms of the Milnor number of the ICIS,
the topological invariants of P and an intersection number.
Theorem A contains as particular cases the Teissier [36] restriction formula,
the Lê attaching formula (in the case of dim Sing V(f) 1) and the Iomdin-Lê
formula. This is not surprising, since in this case the polar curve technique is
replaced by a method based on the properties of the discriminant space of an
ICIS. During the proof, when comparing the Milnor fiber of f with its part
contained in a "good representative" of the ICIS we are using a technical lemma
of Iomdin [13] (Section (3.1)).
The most important topological invariant of isolated singularities is the Seifert
form. In fact, for n 3 it is an absolute invariant in virtue of the classification
theorem of simple spinnable structures on S2n+1, proved independently by Kato
and Durfee. Unfortunately, in general, it is very hard to calculate it. For plane
curve singularities it was calculated by A’Campo [1] and Gusein-Zade [ 11,12].
But for n 2 even in the simpler (but very important) case of quasihomogeneous isolated singularities we do not know this invariant. Moreover, it
appears that even the computation of weaker invariants, such as the intersection
form and the monodromy action is still a very hard problem. Concerning the
zeta function there are some general results: A’Campo [2] determined it in terms
of invariants of the resolution, Varchenko [38] in terms of the Newton diagram,
while Milnor and Orlik [22] calculated it for quasihomogeneous isolated
=
singularities.
calculate (Theorem B) the zeta function of f
P(h, g) in
terms of the zeta function of P (which is well-known), the monodromy
In Section
(2.2)
we
=
65
representation and the singular monodromies of the ICIS (the latter may be
using the vertical monodromies of the ICIS). In the case when the
monodromy representation has an abelian image this formula becomes simpler
and is related to the Alexander polynomial of the link S3~(DP-1(0))S3~
(Theorem C).
In the class of all singularities, the isolated singularities play a distinguished
role. Similarly, in our class of germs, there is a distinguished class: those germs
having the property DnP-1(0)={0}. Any "bad" germ f~=P~(h,g) (i.e.
D n P-1~(0) =1= {0}) can be approximated by a series of distinguished singularities
fk = Pk(h, g). This is similar to the case of Arnold’s series, which approximate
some non-isolated singularities, and with the case of lomdin’s series or the
topological series of plane curve singularities introduced by Schrauwen [26].
The best approximation holds for "topologically trivial series", when the
topological type of Pk and P 00 agree, but their mutual position with respect to D
varies. In this case we calculate the zeta function (in particular the Eulercharacteristic of the Milnor fiber) of fk in terms of the zeta function of f~ and the
vertical and horizontal monodromies of the ICIS (Theorem D). In particular for
the lomdin’s series we recover Siersma’s result [33], which was, in fact, our
starting point in the study of series of singularities.
The polar filtration method used by Siersma is replaced here by the technique
of fibered multilinks (the decomposition of the complement space along the
separating tori, determined by splice decomposition).
In Section (2.4) we prove that the complement X of a projective hypersurface
given by a homogeneous polynomial f P(h, g) is (n-1)-equivalent to the
Eilenberg MacLane space K(03C01(X), 1). If f is "distinguished" then we have a nequivalence, and using this correspondence we calculate the homology groups of
X in terms of 03C01(X) which is also computed. This result may be regarded as a
generalization of a theorem of Kato and Matsumoto [15, Proposition 4.2]. In
particular this procedure can be used for a large class of singular plane curves in
order to compute the fundamental group of the complement space.
In (2.5) we consider the finite I-codimension germs in the sense of Pellikaan
[23, 24], where I = (hk-1) (k &#x3E; 2), and h is a germ with an isolated singularity at
the origin. By our Criterion the germs with finite I-codimension are of the form
f = hkg, with (h, g) ICIS and g having an isolated singularity. Our general
results may be applied to this special case. For example the Milnor fiber is a
bouquet of spheres Sl v (03BCnSn) where y,, k(03BC(h, g) + Jl(h)) + Jl(h, g) + p(g).
For the theory of complete intersections we refer the reader to the book of
Looijenga [20], for singularity theory to the monograph Arnold-Gusein-ZadeVarchenko [3], for link theory to [9], for Section 2.5 to the thesis of Pellikaan
[23] and for the proof of Theorem C to the paper by Fox [10] about the free
calculated
=
=
differential calculus.
66
2. The main results
2.0. Preliminaries and notations
Throughout this paper we shall denote by O the local ring of germs of
analytic functions f: (Cn+1, 0)~(C, 0) and by m its maximal ideal. v(a) denotes
the germ of the zero-set of an ideal ce c (9. If 03B1 is an ideal in O such that Y(a) is a
curve and f~O such that V(03B1) V(f) = {0} then we use the (algebraic)
intersection number (a, f)0 given by dim(coker) - dim(ker) of the map (9lot f+ (9lot
(multiplication by f). The following lemma is well-known:
2.0.1.
LEMMA.
If a is an ideal and f ~O, then we shall denote by al the ideal generated by a in
the ring of fractions of (9 with respect to the multiplicative System {fn}n0. Let
03B3f(03B1) denote the ideal a f n (9 in (9 and 03C3f(03B1) the intersection of the remaining
isolated primary components of a.
2.0.2. Let h, g~O be such that the pair (h,g):(Cn+1,0)~(C2,O) defines an
isolated complete intersection singularity (this will be abbreviated in the sequel
by ICIS). Recall that the Milnor fiber F. of u = (h, g) is a bouquet of Jlu Jl(h, g)
spheres of dimension n-1. (If n = 1, then y(h, g) = (h, g)o - 1).
Let c be the ideal in (9 generated by the 2 x 2 minors in the jacobian matrix of
(h, g). Obviously, the zero-set V(c) of the ideal c is precisely the set of critical
points of the map (h, g). Let D be the reduced discriminant locus u(V(c)) and
D1,...,Ds its irreducible components. Let 0393i1,..., riti be the irreducible
components of u-1(Di) (i=1,...,s) and dij the topological degree of the
branched covering 0393ij~Di (i = 1,..., s; j = 1,..., ti). If z E 0393ij - {0} we denote by
03BCij(z) (resp. Fij(z)), the Milnor number (resp. the Milnor fiber) of the ICIS
(h, g) : (Cn+1, z) ~ (C2, (h, z)(z)). Note that 03BCij(z) does not depend on the choice of
=
z on
Let
ciently
{0}.
D~ = {(c, d) ~ C2 : |c|2 + |d|2 ~2}
small
one
defines the
and
aD" = S3~, * ~ S3~ - D.
S3~ - D is a deformation retract in D" - D, p can
03C01(S3~ - D, *) ~ Aut(Hn-1 (Fu, C)) (still denoted by p).
The discriminant locus defines a link in S’ with
Since
For ’1 suffi-
monodromy representation
be identified with
link components
67
KDi = Di S3~.
Kij = (u|0393ij)-1(KDi)0393ij for each i and j
and consider Tij a
small closed tubular neighbourhood of Kij(in X C"+ 1, where (h, g): ~~D~ is
an "excellent representative" of the ICIS [20]). Let N(KP) be a closed tubular
neighbourhood of Ki in S3~ such that for all (c, d) E N(KDi) the fiber u - ’(c, d) meets
transversely the boundary Tij. We shall denote by Mi and Li an oriented
"topologically standard" meridian and longitude of the link component KP on
ôN(K?), which are determined up to isotropy by the homology and linking
relations
Set
c
(c, d) E KD,
fiber u-1(c, d) = FSingi has 03A3tij=1 dij singular
of
the corresponding Milnor fibers is
disjoint
Therefore we have an action of
03C01(~N(KDi)) Z x Z (generated by the class of Mi and Li) on
which preserves the subspaces
Fij = u-1(c, d) n Tij for each j. In particular
Mi resp. Li defines the meridian resp. the longitudinal monodromy
If
then the
The
points.
singular
union
dijFij
=
Udij
hMi, hLi: Hn-1(
hMi
=
hy O
and after
a
...
O
h(ti)Mi, hLi
=
with decomposition
h~
suitable choice of base of
...
~ h(ti)Li
Et) dij Hn-1(Fij, C)
M ij, Lij: Hn-1(Fij, C) (i 1,..., s ; j 1,..., ti).
Obviously, we have by the commutativity of
=
=
Note that this construction can be regarded as a particular case of the
monodromy of local systems [6]. Mij is in fact the horizontal monodromy of a
transversal section, Lij is the vertical monodromy of the transversal local system.
(See also [33].)
By Thom’s first isotopy lemma ([41]), over KDi u is a topologically locally
trivial map with fiber Fs’°g (which is also a bouquet of (n -1)-spheres), thus
We take
defining a singular monodromy
Hn-1(FSingi, C) = 0 then 0394Singi(03BB)=1).
68
By
a
Mayer-Vietoris argument for all
(in particular the right-hand
Let P: (C2, 0)~(C, 0)
term is
ai E Z
we
have
independent of ai E Z).
analytic germ in two variables, and let
Pi Pmr
prime decomposition of P. Then the Milnor fiber Qp of P
has 03BC0 g.c.d. (ml,.. -, mr) connected components and each of them has the
homotopy type of a bouquet of pp circles.
If D is the reduced discriminant locus of the ICIS (h, g) (see (2.0.2)) there exists
PD : (C2, 0) ~ (C, 0) such that V(pD) D. Let Pi , ... , P, be the common compo...
nents of P and PD, i.e., P PT1
pm-; pD p@ ... PtPD1 pD
2.0.3.
P
be
an
be the
...
=
=
=
=
=
...
2.0.4. Let us define f ~ O by f
P(h, g), where (h, g) is an ICIS and P is as in
(2.0.3). We use the notations Irf V(03B3f(c)), 03A3f = V(03C3f(c)). Recall, that in this case
(91c is a 1-dimensional Cohen-Macaulay ring, and in particular V(c) has no
embedded components at the origin. 0393f(resp. 03A3f) consists of the components of
V(c) that are not (resp. that are) contained in V(f). In particular, if D and V(P)
have no common irreducible components, then 03A3f 0.
=
=
=
2.0.5. If a* E
Aut(H*(X, C)), we define the
Hqdet(1 - 03BBaq)(-1)q+ 1. If f~O, then the
03BE(hf,*)(03BB),
hf: F.
The
following
LEMMA
Then
where
[33].
hf,*:H*(F,C)
zeta function of a* by 03BE(a*)(03BB)=
zeta function of f is
(f(î) =
is induced
by the geometric monodromy
lemma will be very useful:
Let
det(I - ÂM) = det(I - AkM1 M2 ...
Some notations:
Mk).
69
2.1. The
homotopy type of the Milnor fiber
With the above notations
we
have the
following
THEOREM A.
(a) The Milnor fiber F of f P(h, g) has the homotopy type of a space obtained
from the total space of a fiber bundle with fiber Fu and base space Qp by attaching
(y f(c), f)o cells of dimension n.
In particular the natural map F ~ U,40 K(nl(F), 1) is an (n -1)-equivalence (the
latter space is a disjoint union of 03BC0 copies of the Eilenberg-MacLane space), and
the Euler-characteristic of F is :
=
QS, F has the homotopy type of 03BC0 copies of a bouquet of spheres
(03BCn S"), where
(b) If Y-f
(03BC1 SI)
V
=
Moreover, the protection
the map
The
U Ilo Nil 1 Sl
can be
identified to
u: F ~ QP.
proof of Theorem
A will be
given
in
(3.2).
2.1.1. COROLLARY. Let
If n
2 +
03B5(f), then 03C01(u): 7Ti(F)
~
03C01(QP) is
an
isomor phism.
Theorem A contains various well known formulae
as
particular
cases.
EXAMPLES
2.1.2. Let (h, g) be
Assume that P(c, d)
our formula in this
as
=
in Theorem
c; in
case
an
IS
h. Since h is
an
A, with h~O
particular f
=
(isolated singularity).
IS, Y-f 0. Therefore
=
becomes:
(the Milnor-number formula of the ICIS,
see
for
example [20] p. 77).
70
p.
If, moreover, g-1(o) is
317]) formula
a
generic hyperplane, then
we
get Teissier’s ([36,
2.1.3. Let h~O be such that dim Sing V(h)
1, and assume that (h, g) is an ICIS
(the latter condition is automatically fulfilled if g is a generic linear form).
Assume again that P(c, d) c. Then the Milnor fiber F,, of h is (n - 2)-connected
and we have the following relation between the Betti numbers:
=
=
If
is a generic hyperplane,
with
dim Sing J1: h) = 1).
germs
V(g)
we
get the "Lê attaching formula" [16] (for
(h, g) be an ICIS and take P(c, d) = c + dk (k 1), i.e. f = h + g’. Again
Fh+gk Fu{(03B3f, h + gk)o cells of dim n}. Since the curves V(c + dk) (k &#x3E;, 1) have
no irreducible components in common, there exists ko such that if k k0, D and
V(c + dk) have no irreducible components in common, in particular E f 0 and
c = y f (c). In this case h + gk is an IS. By Lemma (2.0.1) and Theorem A:
2.1.4. Let
=
=
If g is generic linear form, then this is exactly the Iomdin-Lê formula [14, 18].
(h, g) be an ICIS such that h and g are IS. Then for Pk,l(c, d) ckdl
(k 1, 1 a 1), 03A3f 0. The numerical invariants of the bouquet are: po g.c.d.
(k, 1), pi= 03BCP 1 and (using (2.1.1 ))
2.1.5. Let
=
=
=
=
2.1.6. Let h(z) zn+1 and g(z) = ~(z1,..., zn), where 9: (C ", 0) ~ (C, 0) defines an
IS. Then by a theorem of Sakamoto [25] the Milnor fiber F of f = hkg may be
regarded as the total space of a fiber bundle over S 1 with fiber the Milnor fiber of
ç (and monodromy mk~). Thus if n &#x3E; 2, 03C0n-1(F) Z4("). Hence our assumption in
Theorem A(b) is necessary.
=
=
2.1.7. Let ( f, 0) be an ICIS such that f + 0 is
o is greater than dimc O/03B3f(c) + (f). Then
an
IS. Assume that the order of
71
This corollary gives an (partial) answer to
reformulated by Siersma [23, p. 115].
2.2. The
a
question
by Wall,
raised
as
zeta function of the monodromy
determine the zeta function of the monodromy of the germ
(P’)40 and f ’ P’(h, g), then we have the following relations
f P(h, g).
between the corresponding Milnor fibers and zeta functions:
In this section
we
If P
=
=
=
Therefore, we may assume (with no loss of generality) that F f is connected, i.e.,
g.c.d. (ml,..., mr)
2.2.1. LEMMA.
=
1.
If ~
is
sufficiently small,
then
locally trivial fibrations of pairs of spaces. Moreover, the fibrations
isomorphic.
are
Proof..
This is
a
are
fiber
small, the fibers ~P- 1(e203C0i03B8) in
transversely the link components
consequence of the fact that for ri
S’-P-’(0) (resp. P-1(w), w~~D103B4) meet
s3 n D (resp. D) (to see this use the Curve
prove the
isomorphism we construct
relative case [21, pp. 52-53].
a
Selection lemma
vector field
2.2.2. REMARK. A fiber of çp can be identified with
of the multilink corresponding to the link:
[20]). In order to
similarly as in the nona
minimal Seifert surface
with
multiplicities (ml, ... , m,, 0,..., 0) ([9], p. 34).
(2.2.1) also follows from the fact that L(m, 0) is a fibered multilink.
Let * be a point of F~ -D(F~ = ~-1P(1)), then by (2.2.1) we have the exact
In fact
sequence of groups:
By the inclusion
G
=
S3~ - P-1(0) D S3~ - D, A = H"-’(F., C)
03C01(S3~ -P-1(0) ~ D, *)-module (induced by p), and by i * a
module.
H
=
becomes
a
03C01(F~ - D, *)-
72
1. Then the maps 03C1g 03C1(g): A ~ A and cg: H - H,
automorphism of the exact sequence (of C-vector
Let g E G such that ~*(g)
h ~ g-1 · h · g
induce
an
=
=
spaces):
Let
us
define
where
By [27,
g. Hence
p.
116] the automorphisms g*0 and g*1 do not depend on the choice of
is well defined by the representation p and the exact sequence
(,,,,
(2.2.3).
2.2.5. REMARK. When 1:f = 0 then g can be chosen such that the image of g
in 03C01(S3~ - D, *) is trivial, hence 03C1g 1A.
The zeta function of P is 03BEP(03BB) 0394P(03BB)/(1 - Â), where 0394P(03BB) is the characteristic
polynomial of the monodromy of P (on H1(Qp, C)). With this notation we have
the following
=
=
THEOREM B. Let
where
f
=
P(h, g) be
as in Theorem A. Then
(a)
kj = (P, Pj’)O(= * (Qp n Dj)) (and the product is taken over the components of
are not contained in P-1(0)).
D which
(b) Suppose Y.,
(the product
=
is taken
QS and n
over
2. Then
the all components
of D).
73
The
proof will
be
given
in
(3.3).
free group with
generators,
be
identified
with
A03BC(P,D)
by 03B4~(03B4(b1),...,
say b1,...,b03BC(P,D), Der(H, A)
03B4(b03BC(P,D))). The corresponding matrix [gDer]: A03BC(P,D)~A03BC(P,D) may be regarded as
the "Jacobian matrix of (G, H) at p" [10]. More precisely:
Let us consider the fôllowing homomorphisms of rings:
2.2.6. REMARK. Since H is
a
can
’t, 1*: Z[H]
- Z[G] induced by i*
p : Z[G] ~ Z[Aut A] induced by
s:
p
Z[Aut A] ~ EndcA given by s(Li ci[aijk]jk) = [Li ciaijk]jk
and the derivations
ring of a group G).
cg(bi) by wi. Then using the derivation
matrix constructed by block-matrices
(where Z[G] denotes the
We denote the words
that
[gDer]
is in fact the
group
rule
we
obtain
Therefore, when we know explicitly the action cg:H ~ H (i.e. the extension
(2.2.3)), and the representation p, the 03BEP,D follows explicitly from the above
identification.
2.2.7. The above considerations can be made more precise if imp is an abelian
group, because in this case the Jacobian matrix of (G, H) at p may be factorized
by the Alexander matrix of (G, H).
Suppose that we have the following commutative diagram:
where x
is the abelianizer.
Obviously,
in this
case we
also have the
following
74
commutative
diagram:
We can choose the base elements ei (0, ... , 1,..., 0) of Gab Zr + S -by taking
topological standard meridians in tubular neighbourhoods of the link components of L (see (2.2.2)). For each i =1, ... , r + s - t we have an automorphism
Ei = Pab(ei) E Aut A (Et+1
Er 1), and multiplicities mi(mr+1 = ···
mr+s-t = 0). The ring group Z[Gab] is Z[03BB1,03BB-11 ..., 03BBr+s-t, 03BB-1r+s-t] where Ai
=
=
..’
=
=
=
=
correspond to ei.
Let 0394(03BB1,..., 03BBr+s-t) (resp. 0394(03BB)) be the Alexander polynomial of the link
r+s-t2 (resp. r+s-t=1) in Z[Gab].
THEOREM C. Ifimp is
an
abelian
group,
L if
then
We recall that A is
± 03BBi11 ···
The
well-defined only up to multiplication by monomials
therefore the above equality is to be taken modulo these monomials.
proof is given
in
(3.4).
2.2.8. REMARK. A particular
03C01(S3~-D, *) itself is abelian. This
is singular of type A1.
case when imp is abelian occurs when
happens exactly when either D is smooth or D
2.2.9. REMARK. Consider the case when D is smooth, V(c) is irreducible and
the Milnor number 03BC(z) of the ICIS (h, g): (Cn+1, z) ~ (C2,(h, g)(z)) is constant for
z e V(c) (in a neighbourhood of the origin). Then Fi’ng is homeomorphic to the
1. Therefore
central singular fiber, hence it is contractible, thus 0394Sing1(03BB)
=
75
where
2.2.10. EXAMPLE. Take h=zn+1 and g(z) = g(z1,...,zn), where g:(Cn, 0) ~
(C, 0) defines an IS. Then D {d = 0}. Let P PT1... Pmrr such that 03A3f 0. Then
=
=
=
then 0394f,n(03BB) = det(1-03BBkhk-1g).
We note that
2.3. Series
0394*
can
be
immediately
calculated
using
the
EN-diagram [9].
of singularities
Let {Pk}kk0 be a series of plane curve singularities, such that
(i) P-1k(0) and D have no common components
(ii) limk~~ multo(P 00 - Pk) 00 (where multo(P) denotes the multiplicity
of P).
=
By Theorem A
we
get for fk
=
Pk(h, g)
2.3.3. EXAMPLE. Consider the
following series
of
singularities
76
Then
2.3.4. Series of singularities appear naturally in the classification of germs. The
Arnold’s series.of isolated singularities are related to non-isolated singularities,
by "approximating" them. This is the main motivation for Schrauwen’s [26]
definition of topological series of isolated plane curve singularities: the topological series belonging to P 00 consists of all topological types of isolated
singularities whose links arise as the splice of the link of P 00 with some other
links.
Similarly, in our case, when P-1~(0) and D do have common irreducible
components, we want to approximate the germ f 00 by germs fk P,(h, g) such
that fk is a "nice germ" in our class, i.e. 03A3fk 0. The best approximation of f~
with fk holds when X(Pk) X(P (0) (see formula (2.3.2)), i.e. when the topological
type of P~ and Pk agree, and only the mutual position of P-1~(0) and Pk 1(0) with
respect to D varies.
=
=
=
We represent it
by the EN-diagram:
i
given from the Newton pairs (pij, qij)j by the formula ai1
+
+ 1+ pijpij+1aij if j 2.
Consider the following developments for i =1, ... ,t :
where aij
a; 1 = q;
with
are
EN-diagram
where ai = qi + psas.
=
qi1,
77
Thus, the link component KDi(i=1,..., t) is replaced by K’i, a cable on KDi,
which will follow KDi around 1-times in the longitudinal direction and ai-times in
the meridian direction.
2.3.5. DEFINITION. For
{ai}i=1.....,t,
all
pmr
...
and
with
we
ai&#x3E;pisais,
say that
Pa
we
is
a
define
"topolog-
ical trivial series" belonging to P~.
If the schematic EN-diagram of the multilink
is the
following:
then the schematic
Thus
EN-diagram
of i
KD1,..., KD remains as "virtual components" of F(P,,), P, and
common
components and X(P a)
r(Pa) is
~(P~).
obtained from 0393(P~) by splicing with t (trivial)
D have
=
torus links:
no
78
By splice condition m’i =
mjt(Ki, Kj) (hence it is an invariant of 0393(P~)
and by an easy computation it may be calculated from the EN-diagram, see
p. 84 [9]). Obviously, the algebraicity condition is fulfilled if ai is sufficiently
large (ai &#x3E; pisais).
With this notation
THEOREM D.
1n
we
If a » 0
have the
then for
following
fa
=
Pa(h, g):
particular
The
proof is given
2.3.6. EXAMPLE
in
(3.5).
(lomdin’s series)
(t = 1, m1 = 1, m’ = 0, a1 = q).
ICIS such that dim Singh -1(0) 1. Thus {c 01 c D. The
series fq = h + gq approximate the "bad" singularity f~ h.
If g is a general linear form, then we obtain Sierma’s result [33].
Let
(h, g)
be
an
=
=
=
2.4. The
complement of a projective hypersurface
be homogeneous polynomials in Cn+1 (n 2) of degree dh 1
and dg 1, such that (h, g) is an ICIS. Let di d,/g.c.d. (dh, dg), d2 d,/g.c.d.
(dh, dg). Let P be a weighted homogeneous polynomial in two variables with
weights dl and d2 and degree d. Then f P(h, g) is a homogeneous polynomial
with degree d f d’ (dh, dg).
Since we want to study the complement space
2.4.1. Let
h, g
=
=
=
=
79
may assume in this section, that P = Pred
It is easy to see that the Milnor fiber F
we
=
covering
over
the
= P1 ··· Pr.
f-1(1) is the
total space of a ZdfTheorem A we
complement X = Pn - [V(f)]. Therefore, by
have:
2.4.2. The natural map X
for the definition of 03B5(f).)
~ K(03C01(X, *), 1) is an (n-03B5(f))-equivalence. (See 2.1.1
2.4.3. The aim of this section is to determine the group 03C01(X, *).
The fiber bundle
with fiber C* has the
E = Cn+1-V(f)Pn-[V(f)]
homotopy
exact sequence
The exact sequence of the fiber bundles EC* with fiber F and
EP = C2 - V(P) pC* with fiber Q p are related by the diagram:
Since ul is an isomorphism for n 2 + 03B5(f) (by (2.1.1)) we obtain that u2* is an
isomorphism. Consequently 03C01(X, *) = 03C01(Ep, *)IN’, where N’* is the normal
subgroup generated by the class of the loop t (h(e203C0it. *), g(e 21rit*))
(e203C0idhth(*), e203C0idgtgt(*)), t ~ [0, 1].
=
complement space {(c, d): |c|2 + Idl2 = 11 _ V(P) can be identified with
E(P) = 03A3 (d2, d1) - U Si, where 03A3(d2, d1) is the Seifert manifold with Seifert
invariants d2 and dl [9], and singular orbits S, and S2 (corresponding to c = 0
and d = 0), and U Si is a disjoint union of S1-orbits such that
2.4.4. The
(i) SI U Si c is a component of P
(ii) S2
component of P
(iii) U Si contains r’ general orbits «P has r’ irreducible factors different from
c
and d.
Let 03C01(E(P), *) with a base point * on a general S1-orbit S* , and let N*
be the normal subgroup generated by the class of the loop which follows the Slorbit S* in (dh, d,)-times. Then 03C01(E(P), *)/N* = 03C01(Ep, *)IN’ by the above
identification.
Our results can be summarized in the following
2.4.5. PROPOSITION.
If n 2 + e(f ), then there exists a (n - e(f »equivalence X ~ K(03C01(E(P), *)IN*, 1). In particular 03C01(X) depends only on the
numbers (dh, dg) and r’, and on the fact that c resp. d is or is not a component of P.
80
This result can be
Matsumoto [15].
regarded as a generalization of Proposition 4.2 of Kato and
These cases may be compared to a general result due to Lê and Saito [19]
which says that 03C01(X) is abelian (hence 03C01(X) H1(X, Z)) provided f has normal
crossing in codimension 1.
=
(b) P = c3 + d2,
03C01(Ep, *) = .
and
n1(X) = (x, y:
xy n1 (X) = (a, b: a2 b3, b3(dh,d,) = 1).
b
=
by substitution
a=xyx,
=
particular if we take (h, g) = (x2 + y2, x3 + z3) then we
example f (X2 + y2)3 + (x3 +Z3)2 with n1 Z2 * Z3 [40].
In
=
obtain the Zariski
=
2.4.7. Suppose that LI
0, hence e(f ) 0. The homotopy groups and the
Euler characteristic of X can be determined by the Zd f-covering, and the
homology groups Hq(X, Z) 0 q n - 1 by the n-equivalence (2.4.5). Since X is
a n-dimensional Stein space, the only non-trivial remaining group, Hn(X, Z), is
free and may be determined by an Euler-characteristic argument.
=
2.5. Germs with finite
=
(h)k-1-codimension
2.5.1. Recall that (9 is the local ring of germs of analytic functions of (n + 1)variables and m is its maximal ideal. For an analytic germ we denote by J f its
Jacobian ideal, i.e. Jf = (~1f,..., ~n+1f)O. If 7 is an ideal in (9, then its primitive
ideal f I is defined as {f E lP(f) + J f c I} [23]. Let -9 be the group of germs of
local analytic isomorphisms h :(Cn 1 1, 0) ~ (Cn+ 1, 0) and consider the subgroup
DI = {h~D|h*(I)=I}. Since DI acts on I, for germs f ~I we can define the
tangent space 03C4I(f) = {03BE(f)|03BE E TDI} (where TDI = {03BE E m Derc : 03BE(I) I}) of the
orbit of f under DI. Similarly, the sub pseudo group DI,e of the pseudo group
De of local analytic isomorphisms H:(Cn+1, 0) ~ Cn+11 determines the extended
tangent space
(where TDI,e= {03BE~D
of f.
The
1-codimension,
resp. the extended I-codimension
of f
is defined
by
81
2.5.2. Consider h ~ m with an isolated critical point at the origin and let
thus if
I = (h)k-1 (k 2). Then by an easy computation we obtain
f ~ I then f = hk. g. It is easy to see, that if g has only isolated singularities then
the critical locus of f is V(h).
The purpose of this section is to determine the topological type of the Milnor
fiber of germs f
hk. g with finite I-codimension.
that I=(h)k,
=
2.5.3. If 9 ft m, then the Milnor fiber of f is diffeomorphic to k (disjoint) copies of
the Milnor fiber of h. Note that even in this simple case, in general cI,e(f) ~ 0. By
[23, p. 37] 03C4I,e((h)k) = Jhk(h)k, i.e. cI,e03BC(k) - 03C4(h) - 03C4(h) (03C4(h) = Tjurina number of
h), hence cI,e(hk) 0 if and only if h is weighted homogeneous.
Therefore the DI-classification is very complicated in general, for example it
may happen that we have no DI-simple germs. (See also [7].)
=
2.5.4. For germs g ~ m
we
have the
following
= hkg ~ I
FINITE I-CODIMENSION CRITERION. Let f
such that
Then cI(f)
oo if and only if the following two conditions hold:
(i) g defines an IS
(ii) (h, g) defines an
f ~ m.
ICIS.
Proof. Suppose that cj(f) is finite. Since 03BE(h)/h~O, for 03BE ~ TDI, we can
consider the ideal /3 {kg03BE(h)/h + 03BE(g)|03BE E TDI}. From the isomorphism
I/03C4I(f) ~ (9jP we obtain that /3 has finite C-codimension. Obviously
fi c (g) + Jg hence g is an IS. Let c be as in (2.0.2). Then the zero set of the ideal
c + (g) + (h) is contained in the zero set of the ideal fi, hence (h, g) is an ICIS.
Conversely, if the pair (h, g) is an ICIS and g is an IS then if we denote by J the
ideal generated by kgih + hô;g, i =1, ... , n+1 in 6, then J + c has finite Ccodimension. (To see this suppose that V(J + c) ~ {0} and use the Curve
selection lemma). Considering the derivations 03BEi = h~i and
we obtain that /3e
{kg03BE(h)/h + 03BE(g)|03BE E TDI,e} c + J, hence cj,,,(f) is finite.
=
=
2.5.5. COROLLARY. If f = hkg has finite
Milnor fiber of f is a bouquet of spheres SI
pn
=
g E Wl, then the
NJJn Sn), where
k(iÀ(h, g) + IÀ(h» + Jl(h, g) + p(g).
Proof.
Use
(2.5.4)
and
2.5.6. If we consider (h,
cI(f)
=
(2.1.5).
g) as in (2.1.6), then the pair (zn+1, ç) defines an ICIS, but
oo.
2.5.7. EXAMPLE. Let
h(z)=zn+1
homotopic
and
g(z)=~(z1,...,zn)+zn+103C8(z).
If
~ m,
then f Éi zkn+1zn (in the latter case the Milnor fiber
then f
is
(h)k-1-codimension,
v
to
S1).
In both
cases
cI,e(f)
=
0.
82
If ç E m2, then cI(f)
and in this case
oo
if and
only if ç : (C", 0)
-
(C, 0) and g E
U have
only IS
2.5.9. EXAMPLE. Let h be a weighted homogeneous germ of degree dh and
weights w1,...,wn+1. Then TDI,e is generated by the Euler derivation 03BEe =
[5]. If g is also
03A3iwizi~i and the "trivial derivations"
with
same
the
weights w1,..., w" +1 and degree dg, then
weighted homogeneous
/3e c + (g), therefore by [20, p. 77] we get again:
=
Obviously,
when
h~m2,
2.5.10. Let n = 1, k = 2 and
results [29] (for n = 1), i.e.
then
hence
Çe’
h(zI, z2) = z2.
Then
our
cI,e(f)
=
cI(f).
results agree with Siersma’s
because
2.5.11. Consider now the case n = 2 and k = 1, i.e. f = gh. Then 03A3f = Y(g) n V(h)
is a 1-dimensional ICIS, hence we can compare our result with [30,31].
Obviously, the Hessian number ô(f ) vanishes, =tt= D 00 0. By the main theorem
of [31], case B, F ~ S1 v NJl2 S2), where the number of 2-spheres in the
bouquet is p(h,g)+*Al (#=A1=j(f)=dimc(h,g)/Jf). By our result # A1=
=
p(h)
+
Jl(g)
+
03BC(h, g).
2.5.12. The following (general) statement was formulated as an (erroneous)
conjecture in the first version of this paper. Thanks are due to the referee for
correcting it and hinting at a way of how to prove it.
If f hk·g has finite (h)k-1-codimension, then
=
The proof can be given with a method as in a paper of Greuel and Looijenga:
The dimension of smoothing components, Duke Math. J. 52 (1985), 263-272.
83
3. Proofs
3.1. A TECHNICAL LEMMA. Let f, h, gE2, all defined in the ball BEo. Let
{0} ~ V c V(f) be a germ of a real analytic set, also defined in BEO such that
(in (3.1)
we
will
We have the
use
the notation
following
Y(h)
for
{z E Beo|h(z)
basic lemma, due
3.1.1. LEMMA. There exists
essentially
0 and
=
01).
to Iomdin
[13].
neighbourhood G (in Beo) of
V - {0} such that for z E B, n G - V(f ) if /31(h. ah + g. Dg) + /32Z Âf - ô f with
03B21, 03B22 ~ R, 03B221+03B222 ~ 0, 03BB ~ C then fil - fl2 0.
Proof. The lemma is a consequence of Lemma 1.6. in op. cit. (In the statement
of 1.6, V is a complex algebraic set, but this difference is not essential in the
proof. See also Lemma 1.5 op. cit. for the real analytic version of the resùlt).
Let M {z E Be| there exist /31’ fl2 ~ R+, fil + fl2 ~ 0, and A E C such that
an
eo
&#x3E; e &#x3E;
a
=
=
3.1.2. LEMMA.
If e
is
sufficiently
small then
(the closure is considered in B,).
Proof. Suppose the contrary. Then there exists an irreducible component W
of M - V(f ) n V(f) such that W 94- V(g) n V(h) and d dim W n V(g) n V(h)
dim W - 1 (all spaces are considered in B03B5 with e sufficiently small). If L is a
generic linear space of codimension d which contains the origin, then the real
analytic set V L n W c V(f) has the property V n V(h) n V(g) = {0} and
=
=
dim V a 1. But the existence of V is in contradiction with lomdin’s lemma.
3.1.3. REMARK. One of the consequences of (3.1.2) is that the Milnor fibers
f-1(~) B03B5 and f-1 (~’) B03B5’, (0 03B5’ 03B5, 0 ~ « B, 0 ~’ « 03B5’) are diffeomorphic.
For other consequences see [13].
3.2.
Proof of Theorem
A
3.2.1. It is easy to see that it is enough to prove the theorem for Jlo = 1.
We choose 03B51 &#x3E; 0 and 03B51 » ~1 &#x3E; 0 sufficiently small such that u =
(h, g)(h,
is a "good representative" for the ICIS (h, g) ([20])
and in D~1, P-1(0) and D have cone structure. Let V(c) c B03B51 u-1(D~) be the
reduced critical locus of u and D u(V(c)) the corresponding (reduced) discrimi-
g)-1(D~1) B03B51 ~ D~1
=
nant locus.
84
Let L be
a
generic line
LD={0}, LP-1(0)={0}.
Let 0
81be
8
(3.2.2). (i) B, is
(ij)
(iii)
a
so
origin: L:{(c,d)~D~1:03B1c+03B2d=0} such that
particular L(h, g) = 03B1h + 03B2g ~ O defines an IS.
at the
In
small that
Milnor-ball for the functions
f
and
L(h, g).
BE c u
Lemma (3.1.2) is valid for E for both sets
Finally
chose 0
we
~1small
~
enough
of maps ( f, h, g) and (L(h, g), h, g).
such that
u: u-1(D~) B03B5 ~ D~ is a "good representative" for u.
(3.2.3). (i)
(ii) D~ is a Milnor-ball for P: (C2, 0) ~ (C, 0).
(iii) The restriction of 03C1: C2 ~ R+, 03C1((c, d)) = |c|2 + |d|2 to D has no critical
value in (0, ~2].
(iv) For all (c, d) E D - P-1(0), P-1(P(c, d)) meets transversely the curve D.
(v) Lemma (2.2.1) is valid for ’1.
If c5
&#x3E;
0 is
sufficiently
small relative to e, il 1
and ’1, then
(i) Fa = f-1(03B4) n Be is the (open) Milnor fiber of f in Be.
(ii) Qâ - P-1(03B4) n D~1, Qa P-1(03B4) n Dl are the (open) Milnor fibers
=
D~1
D~.
Obviously u-1(Q103B4) n Be
of P in
resp.
=
F03B4.
3.2.4. As a first step we prove that for sufficiently small c5, u-1(Qa) n BE is a
deformation retract of Fa. Moreover,
is a deformation retract of f-1(~D103B4)B03B5 by a deformation retract which preserve the
fibers of f.
Proof. The différence of P-1(~D103B4)D~1-P-1(~D)D~ has r connected
components, corresponding to the irreducible components of
P:
Ci03B4
=
{(c, d)~D~1, Pi (c, d)~~D103B4, ~ 03C1((c, d)) ~1},i = 1,...,r.
~u-l(Ci03B4) n Be and c5 small, the projection p., of z and the projection p p of
h - ~h+g·~g on the tangent space Tz(f-1(f(z))) are nonzero by (3.2.2(iii),
For
z
on
this lemma
pz e Rpp . Therefore we can construct a vector field
u-1(Ci03B4) n Be (first locally then by a partition of unity, globally) such that
moreover
by
(a) v’(z) is tangent to f-1(f(z)).
(b) Re v’(z), h êh(z) + g Dg(z»
(c) Revi(z), z 0.
For each
0.
there exists
a
unique smooth path
85
such that
(because p o u(pz(t)) and ~piz(t)~are increasing functions) and f(piz(z))
we can
=
f(z). Then
define the deformation retract
by
3.2.5. Let
similarly
us
as
denote the Milnor fiber of L(h, g) by Lb, L(h, g)
in (3.2.4) we obtain that for ô’ sufficiently small
=
-’(ô’) n Be.
Then
is a
of Lb,.
3.2.6. Let us denote u-l(Qb)nBe by F03B4. Then by construction u:F03B4~Q03B4 is a
smooth fiber bundle over Q03B4 - D with fiber the Milnor fiber F. of the pair (h, g).
A singular fiber u-1((c, d))((c, d) E Q n D) is also a bouquet of (n-1)-spheres and
can be obtained from F. by attaching rank Hn-1(Fu)-rank Hn-1(u-1(c, d» cells
deformation retract
of dimension n.
The cardinal number of the intersection D Q03B4 = {O1,...,ONp} is Np =
(PP, P)o (see (2.0.3) for the notations). It is easy to see that there exists a
subspace Qâ = {bouquet of circles with base point *} in Qa such that Q,, is a
deformation retract of Qa and D n Qa 0. Let Di, i = 1,..., N p be small closed
C~-embedded disks in Qa - Qa with center at the points Oi with radius so small
that they are mutually disjoint and for all i let li be a C°°-embedded interval in Qb
from the base point * ~ Q03B4 to a point 0’ on ôDi such that1 = U i li can be
contracted within itself to *, Q03B4 ~1 can be contracted within itself to Qa and Qô
can be contracted to Q03B4 ~l~ U i Di. Then (by usual arguments, see [28, §7],
[31]) we obtain that #03B4 (hence the Milnor fiber F03B4, too) has the homotopy type
of a space obtained from u-1(Q03B4)~B03B5 by attaching cells of dimension n. But
u-1(Q03B4) ~ Be may be regarded as the total space of a fiber bundle with fiber F.
and base space homotopic to Qp.
In the following we shall count the number of cells.
=
3.2.7. Recall that V(PP i 1, ... , s - t are the irreducible components of u(0393f)
and
Fij is the irreducible decomposition of u-1(V(PDi)). If (c, d) E V(pf),
then u-1(c, d) n Be is obtained from F. by attaching 1 03BCij(z) cells of dimension n
(the sum is taken over z~u-1(c, d)~0393f) [20, pp. 75-76]. Thus each V(PD)
contributes with (PDi, P)o 03A3tij=103BCijdij cells. Therefore the total number of cells is
=
86
On the other hand
(see [36, p. 318]), where n(03B3ijf)) is the length of ON (N = ideal of nilpotent elements
in O/03B3ijf).
In order to prove the
equality J.lij
=
n(03B3ijf), consider the ICIS
Then the reduced critical locus 0393redij,z is smooth and (h, g)z |0393redij,z is an immersion.
Consequently, the rank of dz(h, g) is not zero, hence at least one component of
(h, g) is smooth at z and meets 0393ij transversely at z, hence by [20, 5.11(a)] and
using the above formula, we get pij n(y’j
Consequently, the Milnor fiber Fa has the homotopy type of u-1(Q03B4) n BE with
(03B3f(c), f )o n-cells attached, which proves (A(a)).
=
3.2.8. For the definition of the natural map F ~ UJLO K(03C01(F),
3.2.9. Let
n
=
1. Then
3.2.10. Let n &#x3E; 2 and
that
(a) (3.2.4)
(A(b)) follows from
an
1)
see
[34, p.427].
Euler-characteristic argument.
1:¡ = 0. Fix ô’ as defined in (3.2.5) and take
à
&#x3E;
0
so
small
is valid.
(b)P : {(c, d) ~ D~is :smooth
03B1c
+ 03B2d
fiber bundle.
P-1(~D103B4) n D~ ~ ôD§
(c)
Let
a
Q|03B4| P-1(~D103B4~D~ and Q|03B4| P-1(C-D103B4)~D~.
=
=
By
argument similar to that of [21, pp. 52-53] using moreover (3.2.3) we
n D) is a deformation retract of (Q|03B4|, Q|03B4| D).
obtain that the pair
is a deLifting this deformation retract we obtain that
formation retract of
(see also the argument of [20, 2.D]).
Let *’ be a point on Q03B4 ~ {(c, d) : ac + /3d 03B4’}.
As in (3.2.6) we consider the intersection points {03B1c + /3d ô’l n D, small closed
C ’ -embedded, mutually disjoint disks D’ in {03B1c + 03B2d = 03B4’} with center at these
points and embedded intervals h from *’ to points on ÔD’ such that
an
(Qpl, Qisl
E|03B4| = u-1(Q|03B4|)~B03B5
=
=
(i)
(ii)
(iii)
~il’ican be contracted within itself to *’.
is a deformation retract in D~ ~ {03B1c + 03B2d = 03B4’}.
=
Q|03B4|.
L’
L’
87
Since
u
is
locally trivial
over
D~~{03B1c + 03B2d = 03B4’} - D, u-1(L’)~B03B5
homotopy type of L03B4’~03BC-1(D~)
Since La. is (n-1)-connected (because
morphism, induced by inclusions:
which is
has the
deformation retract in La. by (3.2.5).
L(h, g) is IS, see (3.2.1)), the composed
a
Obviously, since Qa - D is connected, for all
E|03B4| induces a trivial morphism at 03C0n-1.
is trivial.
u-1(*) ~
But this is true also for the inclusion
retract in
*
E
Qa - D the inclusion
u-1(*) E|03B4| because E|03B4| is a deformation
~
E|03B4|.
From
the
homotopy
f -l(aDJ) n Bt 4 ~D103B4
is
exact
with fiber
F03B4,
of
obtain that
sequence
we
the
Milnor
fibration
Bibi
that n 2). By (3.2.4)
is a deformation retract in
a
deformation preserving the fibers of f, hence
is also injective. Looking at the composition
we obtain that
is also trivial.
injective (recall
f-1(~D103B4) ~ B03B5 by
u-1(*) ~ F03B4 ~ E|03B4|
3.2.11. Let us summarize our partial results:
F03B4 has the homotopy type of a space obtained from the total space of the
fiber bundle (T, F*) = (u-1(Q03B4) n BE, F*) ~ (Q03B4, *) by attaching N cells of dimension n such that the morphism (induced by the inclusion) nn - 1(F*) ~ 03C0n-1(F03B4) is
trivial.
- F* is a bouquet of y. spheres of dimension n -1 (with natural base point bo).
Qa is a bouquet of Jlp circles (with base point *).
be the geometric monodromies of the fiber bundle corresponding i = 1,..., 1À,,. Then the total space T can be regarded as a factor set
of a disjoint union ep 1 {i} x F* x [0, 1]/ ~ where (i, x, 0) - ( j, x, 0),
(i, x,1 ) ~ (i, hi(x), 0). We may suppose that hi (bo) bo. Let 1 S1 be the image of
~i{i} {b0} [0, 1] in T. Then T is obtained from (S1) v F * by attaching
03BCP·03BCu cells of dim. n.
Therefore F03B4 has the homotopy type of a space obtained from ( 1 Sl) v F*
by attaching N + 03BCP·03BCu cells of dimension n, such that F* ~ F03B4 induces the
trivial morphism at 03C0n-1. If n 3, then it is immediate that we may assume that
the images of the characteristic maps of all n-cells are in F*, thus
Let
hi: F* ~ F*
to
=
88
the
Obviously,
projection
map F03B4~Q03B4.
NJlp S1) v NJln S") --+VJlP S1can be identified with the
In order to prove the theorem for n
=
2,
we
give
the
following
3.2.12. DEFINITION. A continuous map p: E - B between topological Hausdorff spaces is called a m-quasifibration (0 m ~) if for all points b ~ B and
e ~ p-1(b) the induced maps p* : nq(E, p -1(b), e) ~ 03C0q(B, b) are isomorphisms for
0 q m - 1 and epimorphisms for q m. (For m oo we obtain the usual
definition of quasifibration [8]).
A careful inspection of the proof of Satz 2.2 in [8] shows that the following
holds true:
=
=
3.2.13. CRITERION. Let
(Ui)i~F
be
an
open
covering of B
such that
(a) For each i~F, p: p-1(Ui) ~ Ui is a m-quasifibration.
(b) Each non-void intersection Ui n Uj can be written as an union of elements
in u.
Then p is
a
m-quasifibration.
3.2.14. Consider the map F03B4 ~ Q03B4 and an open covering with small disks of the
space Q03B4: Then 03C01(u-1(b))~03C01(u-1(D)) is an epimorphism for small disks D, and
b ~ D hence 03C0q(u-1(D), u-1(b), e) 0 for q 0, 1. Therefore by (3.2.13) u is a 2is an isomorphism. But
quasifibration, in particular
from the following diagram
=
03C01(Q03B4) is an isomorphism.
Consequently, we may assume that the images of the characteristic maps of all
2-cells are in F*, thus finishing the proof.
we
3.3.
obtain that
03C01(F03B4)
=
Proof of Theorem
3.3.1. We
~
B
keep the notations of (3.2). (For simplicity suppose n &#x3E; 2).
By (2.2.1) we can identify the pairs (Q03B4, Q03B4 -D) and (F~, F~ - D) and the
representations 03C01(Q03B4 - D, *) ~ Aut A and 03C01(F~ - D, *) ~ Aut A (both induced by
P).
By (3.2.4) Ejôj is a deformation retract in f-1(D103B4) n B, and (since P-1(0) and
D have a’cône structure in D~1)E V(c) =
V(c). We have the
commutative
following
diagram.
89
Consequently the monodromy
monodromy of f|E|03B4|.
of
f (over
can
be identified to the
that u is a Thom map over D103B4. Therefore by the Thom’s
second isotopy lemma ([41]) u is locally trivial over able Hence the geometric
monodromy of f |E|03B4| can be regarded as a lifting of the geometric monodromy
of P.
It is easy to
see
Q|03B4|
n D in
neighbourhood of
Qlal. Let
and
u-1(T)~B03B5=F,
T) n B,,. Then
WY
If
and
a.r.
we
introduce
the
new
strata
a.r resp. T,
n 9Bibi
then again by the isotopy lemma we may assume that the geometric monodromy of f |Bibi is a lifting of the geometric monodromy of P which preserves the
subspaces T, ô T, Qlal int T. Consequently
3.3.3. Let T be
a
small closed tubular
=
=
-
where
and
are the corresponding zeta functions of the
restricted to the spaces 1 F03B4, WY ~F03B4 and ôi n -97&#x26;.
3.3.5.
(which
is
monodromy
~ F03B4 is the total space of a fiber bundle over the base space int T
diffeomorphic to Qa - D) and with fiber Fu. By [34, p. 498] there is a
convergent E2 cohomology spectral sequence of bigraded algebras with
Es,t2= Hs(Q03B4 - D, H’(F., C)) and Eoo the bigraded algebra associated to some
filtration of H*(, C), but D is a K(H,1)-space, hence
Hs(Q03B4 - D, Ht(Fu, C)) HS(H, Ht(Fu, C)) which is trivial if t ~ 0, n -1 or s 2.
=
(We recall that HS(free
groups
are:
group,
M)
=
0 whenever s 2
[35, p. 220].) The other
90
= P-D · 03BEP,D, because the monodromy action on
Hq(H, A) q 0,1 can be identified to the action described in diagram (2.2.4). (Here
03BEP-D is the zeta function of the monodromy of P induced on Q03B4 - D.)
Therefore
Eit
=
Es,t~ and
=
3.3.6. Removing the points Qd n D in Q03B4 - D, by the Mayer-vietoris theorem
we obtain that
(P=(p-D’(QnT.«(Q6noT)-1. Since QdnoTis a disjoint union of
circles, (Q6noT = 1; and (Q6nT = (Q6nD’ But QD
{(P, PDi)o = kj points}
and the monodromy can be identified with the cyclic permutation of the points
of these sets, therefore by (2.0.5).
=
Thus
3.3.7.
u-1(Q03B4 n D) is a deformation retract in F03B4 ~
FSingj}
and the
hj,kj o... hj.l
induces
disjoint copies
such that
operator
hSingj.
of
o
Thus
F,
u-1(Q03B4 n D)
geometric monodromy
on
Hn-1(FSingj, C)
the
is
a
cyclic
=
{kj
map:
singular monodromy
by (2.0.5)
Q03B4 n aT
{Sj1 ~ ··· u Sjkj} where Si are smooth circles around the
can be regarded as a
points of D n Qb. The geometric monodromy on F03B4 n
3.3.8.
cyclic
=
map:
91
(h’. J.k) o...oh’. J.1but
Therefore (03BB)
due to the following
=
3.3.9. LEMMA. Let
pr1 p: E ~ S1 (pr,
=
all these zeta functions
are
trivial
ES1 x S’ be a fiber bundle. If we consider the application
first projection) then the zeta function of the monodromy of
trivial, i.e. 03BEpr1p(03BB) = 1.
Proof. The monodromy induces the following automorphism of the Wang
pr1 p is
exact sequence of the fibration
Eo pr2°p) Swith fiber F:
where E0 =(pr1 P)-1(1) F = p-1(1,1).
3.3.10.
Finally
if we summarize:
3.3.11. If 03A3f = , then Hq(F, C)
Theorem A(b) hence (b) is trivial.
3.4.
Proof of Theorem
can
be identified to
Hq(Q(), C)
q
=
0,1 by
C
3.4.1. Let b1,...,bv (v=03BC(P, D)) be a system of generators for the group H, and
1.
let g~G such that 9.(g)
If we consider the element g as a "distinguished generator" [10], then we have
a presentation of the pair (G, g):
=
The Jacobian matrix
of (G, g) is
the
following (see
op.
cit.):
92
If we apply the abelianizer x, then we obtain the "Jacobian matrix at
is the Alexander matrix of the pair (G, g):
x" which
Similarly, if ~(g) = 03A3jajej,
î(g) I1j îjai(X: Z[G] -Z[Gall is induced by X). By [10, 6.3 and 6.4]:
If
~(bi) = 03A3aijej,
then
then
=
Since ~(b-1i(i=1,...,v) and î(g)
determinant of the following matrix
is also
equal
3.4.2. Let
us
to
consider the evaluation map:
given by
This map extends to
3.4.3. Let
us
are
compute
invertible elements in
(mod ±
).
Z[Gab], the
93
If *:Z[G]~Z[Z]
= Z[03BB, 03BB-1] is induced by ~*, then
iP*, hence
Since
Therefore
det(e 0 1(M~)) = det(-1 + 03BB[gDer]) ± 0394Der(03BB).
Since 03B5det Mx = 0394*(03BBmE1,...,03BBm
3.4.4.
that det(e det
MX)
=
we
obtain
det 0394* ·Ag(03BB).
Hence
Therefore, Theorem C follows from the following
3.4.5. LEMMA. det(03B5
Proof. See [4], §12, p.
3.5.
Proof of Theorem
1(M) = det 03B5 det M if M
E Mv
v(Z[Gab])
140, Lemma 2.
D
3.5.1. The fiber of P~ can be identified to a minimal Seifert surface Q~ of the
multilink (S3,
m1K). Let Ni be a tubular
neighbourhood of the link component KD with topological standard meridian
Mi and longitude Li (i =1,...,t). The minimal Seifert surface Q 00 meets the torus
ôNi in di parallel copies of Ki(pi, qi) where Ki(pi) qi) denotes the unique (up to
isotopy) oriented simple closed curve in bNi homologous to piLi + qiMi,
di g.c.d. (mi, m’i), diPi
=
Let
us
=
mi,
diqi = - m’ [9, p. 30].
consider the Seifert links
Ki
with topological standard meridian
Let gi be a tubular neighbourhood of
Seifert surface of the Seifert link
minimal
on
be
a
and
Let
ei
Ûi longitude Li ôNi.
94
Yi. After splicing, Mi identifies to
fiber Qa of Pa to the space
(by
the identification
3.5.2. The Seifert link
Li, Li
to
Mi, Qi n ôNi
to
Q~ ~ ôNl and the
of Q~ - ~ti=1 int Ni to Qa - ~ti=1 int Ni).
i has
a
realization in the Seifert manifold
{(z1, z2) E e2:
|z1|2 + |z2|2 = 1} = 03A3(03B11, 1)
where
Ki = {z1
is
01, K’i = {z2 = 0} and KD is a general S1-orbit, where the Sl-action
given by
The bord
of
=
agi of Ni is given by Iz 11= 03B5 (for 8 small), hence the di parallel copies
diKi(pi, qi)
Let
can
be
given by
the
(z01, z02) ~ ~i-1(1) with |z01|=03B5,
equations:
where
1 is the fibration map of the Seifert manifold.
is given by the équations :
defined by ~(z1, z2)
Then the S1-orbit (e203C0i03C4z01,
=
Therefore the S1-orbit of (z°, z2) meets diKi(pi, qi) in m’i + aimi points. The
is a cyclic permutation of
monodromy action
these points. Thus, we can denote the points by Ti1,...,Tim’i+aimi such that
hi(Tij) Tij+1; j 1,...,m’i + aimi (and
1 = Ti1).
=
3.5.3. The
=
virtual
link
(z11, z12)~~-1i(1), |z11|= 8(1
KDi
+
general S 1-orbit t--+t*(zL zà) where
ro), t ~ C, |t| 1 and ro &#x3E; 0 is sufficiently small.
is
a
=
95
KD meets
the fiber
(p- ’(1) in m’i+ a’mi points given by
points Ti1,...,
These
Tim’i+miai
are
also
the
equations:
circularly permuted by hi (i.e.
hOT’;)
=
Let
T’ in
ai : [0, 1] ~ ~-1i(1), i 1,..., t be an embedded segment between Ti1
~-1i(1) such that 03B11i(0) Tl, 03B11i(1) Ti1,
Tij+1).
=
=
|(03B11i(03C4))1|~(03B5, 03B5(1
+
r0))
if
and
=
1:E(O, 1)
Then
embedded segment between Tij and Tij.
Finally, let Dij be small closed embedded disks with sufficiently small radius
is a deformation
and centers in Tij such that
retract in ~-1i(1)-int 9,, moreover, the deformation retract may be chosen
compatible with the monodromy action.
is
an
Ri
=
3.5.4. Since u is locally trivial over ~ti=1 Ni and Q~ - ~ti=1 int Ni is a
deformation retract in Q~, we obtain that the space u-1(Q~ - int Ni) is a
deformation retract in u-1(Q~) = F~.
is a deformation retract in Qa, and over the
Since
difference space u is locally trivial, we obtain that u-1(Qa) can be replaced by
u-1(Q~)~iu-1(jim03B1~Dij) glued over u-1(jTij). Obviously the monodromy preserves this decomposition, hence
where
03BEi1
03BEi2
By
=
zeta function of the
=
zeta function of the
monodromy induced
monodromy induced
on
on
ju-1(Tij)
Lemma 2.0.5
03BEi1(03BB)
=
(zeta function induced by the Sl-orbit Kf
on
u-1(Ti1))(03BBm’i+miai)
96
Hence
and
by (2.0.5)
we
get formula of Theorem D.
Acknowledgements
1 wish to thank A. Zaharia for many helpful conversations related to the present
paper. Special thanks are due to Professor D. Eisenbud, who provided me with a
copy of his joint book with W. Neumann [9], from which 1 learned some useful
techniques.
References
[1] A’Campo, N., Le groupe de monodromie du deploiement des singularités isolées de courbes
planes I, Math. Ann. (1975) 213, 1-32.
[2] A’Campo, N., La fonction zeta d’une monodromie. Commentarii Mathematici Helvetici (1975)
50, 233-248.
[3] Arnold, V.I., Gausein-Zade, S.M. and Varchenko, A.N., Singularities of differentiable maps,
Vols I and II, Birkhäuser, 1988.
[4] Bourbaki, N., Éléments de mathematiques, Livre II, Algèbre, Chap. 8.
[5] Bruce, J.W. and Roberts, R.M., Critical points of functions on analytic varieties, Topology Vol.
27 No. 1, 57-90 (1988).
[6] Deligne, P., Le formalisme des cycles évanescents, SGA VII2, Exp. XIII, Lecture Notes in Math.
340, 82-115 (1973).
[7] Dimca, A., Function gems defined on isolated hypersurface singularities, Compositio Math. 53
(1984), 245-258.
[8] Dold, A. and Thom, R., Quasifaserungen und unendliche symmetrische producte, Ann. Math.
67 (1958).
[9] Eisenbud, D. and Neumann, W., Three-dimensional link theory and invariants of plane curve
singularities. Ann. of Math. Studies, Princeton Univ. Press, 110 (1985).
[10] Fox, R., Free differential calculus II, Math. Ann. 59(2), March (1954).
[11] Gusein-Zade, S.M., Intersection matrices for certain singularities of functions of two variables,
Funk. Anal. Pril. 8(1) (1974) 11-15.
[12] Gusein-Zade, S.M., Dynkin diagrams of singularities of functions of two variables. Funk. Anal.
Pril. 8(4) (1974) 23-30.
[13] Iomdin, I.N., Local topological properties of complex algebraic sets, Sibirsk. Mat. Z. 15(4)
(1974), 784-805.
[14] Iomdin, I.N., Complex surfaces with a one dimensional set of singularities, Sibirsk. Mat. Z,
15(5) (1974), 1061-1082.
97
M. and Matsumoto, Y., On the connectivity of the Milnor fibre of a holomorphic
at a critical point, Proc. 1973 Tokyo Manifold Confer., 131-136.
Lê Dung Tráng, Calcul du nombre de cycles évanouissants d’une hypersurface complexe, Ann.
Inst. Fourier (Grenoble) 23 (1973) 261-270.
Lê Dung Tráng, Le monodromie n’a pas de points fixes, J. Fac. Sci. Univ. Tokyo Sec. IA.
Math, 22 (1975), 409-427.
Lê Dung Tráng, Ensembles analytiques complexes avec lieu singulier de dimension un (d’apres
I.N. Iomdin). Séminaire sur les singularités, Publ. Math. Univ. Paris VII, p. 87-95 (1980).
Lê Dung Tráng and Saito, K., The local 03C01 of the complement of a hypersurface with normal
crossings in codimension 1 is abelian. Arkiv for Mathematik 22(1) (1984) 1-24.
Looijenga, E.J.N., Isolated singular points on complete intersections, London Math. Soc. Lect.
Note Series 77, Cambridge University Press, 1984.
[15] Kato,
function
[16]
[17]
[18]
[19]
[20]
[21] Milnor, J., Singular points of complex hypersurfaces, Annals of Math. Studies of Math. Studies,
61, Princeton Univ. Press, 1968.
[22] Milnor, J. and Orlik, P., Isolated singularities defined by weighted homogeneous polynomials,
Topology, Vol. 9. pp. 385-393.
[23] Pellikaan, R., Hypersurface singularities and resolutions of Jacobi modules. Thesis Rijksuniversiteit Utrecht, 1985.
[24] Pellikaan, R., Finite determinacy of functions with non-isolated singularities, Proc. London
Math. Soc. (3), 57 (1988), 357-382.
[25] Sakamoto, K., Milnor fiberings and their characteristic maps, Proc. Intern. Conf. on Manifolds
and Related Topics in Topology, Tokyo, 1973, 145-150.
[26] Schrauwen, R.: Topological series of isolated plane curve singularities, Preprint Rijksuniversiteit Utrecht, 1988.
[27] Serre, Jean Pierre, Local fields, Graduate texts in math. 67 (1979).
[28] Siersma, D., Classification and deformation of singularities, Acad. Service, Vinkeveen, 1974.
[29] Siersma, D., Isolated line singularities, Proc. of Symp. in Pure Math., Vol. 40 (1983), Part 2, 485496.
[30] Siersma, D., Hypersurface with singular locus a plane curve and transversal type A1. Preprint
406, Rijksuniversiteit Utrecht (1986).
[31] Siersma, D., Singularities with critical locus a 1-dimensional complete intersection and
transversal type A1, Topology and its applications 27 (1987), 51-73.
[32] Siersma, D., Quasihomogeneous singularities with transversal type A1, Preprint 452,
Rijksunversiteit Utrecht (1987).
[33] Siersma, D., The monodromy of a series of hypersurface singularities, Preprint University
Utrecht (1988).
[34] Spanier, E.H., Algebraic Topology, McGraw-Hill, New York, 1966.
[35] Suzuki, M., Group theory I, Grundlehren der mathematischen Wissenschaften 247.
[36] Teissier, B., Cycles evanescents, sections planes et conditions de Whitney, Astérisque 7 et 8,
1973, 285-362.
[37] de Jong, T., Non-isolated hypersurface singularities, Thesis, Nijmegen, 1988.
[38] Varchenko, A.N., Zeta function of monodromy and Newton’s diagram. Invent. Math. (1976),
37, 253-262.
[39] Zaharia, A., Sur une classe de singularités non-isolées. To appear in Rev. Roum. Math. Pure
Appl.
[40] Zariski, O., On the problem of existence of algebraic functions of two variables possessing a
given branch curve. Amer. J. Math. 51 (1929).
[41] Yung-Chen Lu, Singularity theory and an introduction to catastrophe theory, Universitext,
Springer Verlag (1976).